Microsoft CEO Satya Nadella warns that companies using AI are handing over their most valuable asset

Microsoft CEO Satya Nadella warned that companies adopting artificial intelligence (AI) could be handing over one of their most valuable assets without realising it.

In a post on X, Nadella argued that while businesses pay AI providers for access to powerful models, they may also be sharing proprietary knowledge that helps those models improve over time.

The 'reverse information paradox'


To explain his point, Nadella referred to what economist Kenneth Arrow described as the "Information Paradox".

The idea is that someone selling information cannot fully prove its value without first revealing it. Once the buyer sees that information, they effectively have it for free.

Quoting Arrow, he wrote: "Its value for the purchaser is not known until he has the information, but then he has in effect acquired it without cost."

"AI creates the reverse problem," Nadella said. "In the AI age, the buyer risks giving away knowledge, just in order to use what they bought."

He explained that businesses effectively pay for AI twice. "You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!"

Why company knowledge is at risk

Nadella said that the “information asymmetry” grows over time because AI providers learn more about their customers, while customers have little visibility into what the providers learn.

According to Nadella, resolving this issue requires more than standard data protection. "Models learn from 'exhaust,' the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong," he wrote. In simple words, data users generate while interacting with AI systems, including prompts, responses and corrections.

He added: "Every correction is distilled into institutional know-how. It's the kind of knowledge a competitor could never buy, and the kind that leaks almost imperceptibly: trace by trace, correction by correction, eval by eval."

Who owns the knowledge?

Nadella also pointed to what he described as an irony in today's AI ecosystem.

He said model providers often rely on fair use to train their systems on publicly available data, but at the same time place restrictions on model distillation while retaining the ability to learn from how customers use their products.

If knowledge flows only in one direction, he argued, the value ultimately shifts to the companies that own the AI infrastructure rather than those creating the knowledge.

Quoting Palantir CEO Alex Karp, Nadella said technical customers increasingly want ownership of the means of production.

"The current regime does precisely the transfer Karp and companies fear," he wrote.

How businesses can protect themselves

To address the problem, Nadella said organisations need more than traditional data security measures.

He said enterprises should prioritise control, capability, choice, cost, and compounding by building private evaluation systems, retaining ownership of organisational memory, and creating proprietary learning environments within their own tenant boundaries.

He also called for separating the orchestration layer from any single AI model so companies can maintain flexibility, reduce long-term costs, and preserve the value they create.

"In other words, a company should be able to use a model without giving up the knowledge that makes it unique. That is the reverse information paradox we need to confront," Nadella concluded.